Revolution in AI research: New neural networks imitate human vision!

Revolution in AI research: New neural networks imitate human vision!

On June 26, 2025, a new study on topographic neural networks (All-TNNs) was published in the renowned journal Nature Human Behavior can be found. This research was conducted under the direction of Professor Tim C. Kietzmann at the Institute for Cognitive Science at the University of Osnabrück. The aim of this study is to build a bridge between advanced artificial intelligence (AI) and biological plausibility.

The All-TNNs represent an innovative approach to organizing information. Its principle is based on a two-dimensional replica of the human visual system, similar to the “maps” in the visual cortex. This represents a significant advance because, while traditional convolutional neural networks (CNNs) enable visual feature recognition, they do so in a way that differs from how the human brain actually processes them. Dr. Kietzmann emphasizes that CNNs do not reflect the biological basis of visual processing.

The advantages of all-TNNs

Through the spatially organized selectivity of features on the cortical surface, All-TNNs enable human behavioral patterns to be captured more precisely. These models could revolutionize the understanding of the neural mechanisms behind perception and behavior. Simulations also show that these systems, as physical models, require less energy, which makes them more resource-efficient.

A central feature of All-TNNs is the coordinated work of neighboring neuronal units, which is comparable to natural processes. The challenge of establishing fluid feature selectivity in space represents a central research field. Scientists have already developed promising approaches to optimizing all-TNNs, for example through high-quality image data sets and recurrent connections.

Applications and challenges

The potential applications of all-TNNs range from neuroscience to psychology. Through their ability to simulate human perception, these models could also open new avenues in cognitive psychology and behavioral neuroscience. Their development could also have a significant impact on the design of future AI models.

A crucial aspect of the research is the search for a common language between artificial intelligence and neuroscience. All-TNNs could help overcome the challenges of explainable and understandable AI, which is crucial in today's research landscape.

In the context of applied neuroscience, integrating technology into everyday life remains an important challenge. Neuroadaptive technologies that adapt technical systems to people's abilities and needs offer potential here. The team at Fraunhofer IAO works intensively on the interface between humans and machines in order to develop systems that can adapt to the cognitive and emotional states of users in real time.

In summary, All-TNNs can be viewed as a promising innovation in the field of artificial intelligence and neuroscience. Its development could not only transform the way we think about AI, but also open up new opportunities for research in psychology and other related disciplines. Further information on these exciting topics can be found in the respective research articles from Actu.AI and Fraunhofer IAO.